Add per eval score threshold support to Evalite

[cantemizyurek]

Per-eval scoreThreshold override

Adds the ability to set scoreThreshold per-eval, overriding the global threshold when specified.

Usage

evalite("Eval that requires high percision", {
  data: [...],
  task: ...,
  scoreThreshold: 95, // Requires 95%
});

evalite("Baseline Eval", {
  data: [...],
  task: ...,
  scoreThreshold: 80, // Only requires 80%
});

Behavior

• Per-eval threshold overrides global when set
• Falls back to global threshold when not specified
• If ANY eval fails its threshold, exit code is 1

Watch mode output

When thresholds fail, shows which evals failed:

 FAIL  Score threshold not met. Watching for file changes...
       - High-Quality Eval: 75% < 95%
       - Baseline Eval: 60% < 80%

Solves: #347

[changeset-bot]

🦋 Changeset detected

Latest commit: 73636d9

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

@cantemizyurek is attempting to deploy a commit to the Skill Recordings Team on Vercel.

A member of the Team first needs to authorize it.

[mattpocock]

Keeping this on ice for now, will put this in the post-v1 milestone.